On the classification of
modules over elliptic algebras
Koen De Naeghel, talk s´eminaire d’alg`ebre Institut Henri Poincar´e, Paris
Based on joint work with Michel Van den Bergh. Part of this research is unfinished.
1. Motivation
2. Some preliminaries on noncommutative planes
3. Moduli spaces for gk-2 modules
4. Hilbert series of gk-2 modules
1. Motivation
• First Weyl algebra
A1 = Chx, yi/(yx − xy − 1)
Cannings and Holland, Wilson:
R = { right A1-ideals}/=∼ ←→ a n
Cn
where
Cn={X, Y ∈ Mn(C)| rk(Y X−XY −id) = 1}/Gln(C)
is the n-th Calogero-Moser space
• Berest and Wilson gave a new proof using noncommutative algebraic geometry:
right A1-ideals A1 = H/(z − 1)
H = Chx, y, zi
reflexive graded right
certain objects cohP2q = H-modules
H-modules of rank one
up to finite length where yx − xy = z2
zx = xz, zy = yz
Due to a Theorem of Bondal: Db(coh P2q) RHom P2 q(E,–) −→ ←− – ⊗L∆E Db(mod ∆)
where E = O(2) ⊕ O(1) ⊕ O and ∆ is the quiver
X0
−→ −→X1
0 −→Y0 1 −→Y1 2
Z0
−→ −→Z1
with the relations of ∆ reflecting the defining relations of H: X1Z0 = Z1X0 Y1Z0 = Z1Y0 X1Y0 − Y1X0 = Z1Z0
Under this equivalence:
right A1-ideal ←→ M ∈ mod ∆ for which
dimM = (n, n, n − 1) and
Hom∆(M, p) = Hom∆(p, M ) = 0 ∀p ∈ P1
which is equivalent with
dimM = (n, n, n − 1) and M (Z0), M (Z1) are surjective Using the relations of ∆:
M (X0)M (Z0)−1, M (Y0)M (Z0)−1 !
• There are more algebras inducing a P2q
Interesting class:
A = Artin-Schelter regular algebra in three variables
are determined by geometric data (E, σ, L). Set
R = { reflexive graded right
A-modules of rank one }/=,∼ sh
If E smooth and O(σ) = ∞:
M ∈ R ←→ M ∈ mod ∆ for which
dimM = (n, n, n − 1) and
Hom∆(M, p) = Hom∆(p, M ) = 0 ∀p ∈ E More subtle to handle!
In case A is a Sklyanin algebra i.e.
Skl3(a, b, c) = khx, y, zi/(f1, f2, f3)
where k algebraically closed field char. zero and f1 = ayz + bzy + cx2 f2 = azx + bxz + cy2 f3 = axy + byx + cz2 Then R ←→ a n Dn
Dn smooth affine connected variety of dim 2n D0 = point, D1 = P2 \ E
Remark: Nevins and Stafford: for any Artin-Schelter algebra, without affine part
• Consider (simple) modules over A1 of gk-dimension one.
Determined by Block (1981):
B = localisation of A1 at k[y] \ {0} (PID) The simple A1-modules are
A/(A ∩ Bb) where b ∈ B irreducible (with technical condition on b)
and
k[x] where y acts as y − α = −dxd , α ∈ C
Question: Is there a ’space’ parameterizing these modules?
(simple) right A1 A1 = H/(z − 1)
H = Chx, y, zi (critical) graded right
certain objects cohP2q = H-modules H-modules of gkdim 2 up to finite length where yx − xy = z2 zx = xz, zy = yz Moduli spaces modules of gkdim 1 z-torsionfree
• We work in a general setting
A = Artin-Schelter regular algebra in three variables
E smooth and O(σ) = ∞ Questions:
1. Is there a ’space’ parameterizing (critical) A-modules of gkdim 2? 2. Appearing Hilbert series?
Minimal resolutions?
3. Presentation up to lower gk-dimensional modules?
2. Some preliminaries on nc planes
• Artin-Schelter algebra of dimension 3 is (i) graded k-algebra A = k + A1 + A2 + . . .
global dimension 3
(ii) A has polynomial growth
(iii) A is Gorenstein, i.e. for some l ∈ Z
ExtiA(kA, A) ∼= (
Ak(l) if i = 3,
0 otherwise. • Artin-Schelter: either 3 or 2 variables.
We consider case of 3 variables. Then l = 3 and A is Koszul, i.e.
0 → A(−3) → A(−2)3 → A(−1)3 → A → kA → 0
• They are left and right noetherian domains and
Tails(A) = GrMod(A)/ Tors(A) GrMod(A) π −→ ←− ω Tails(A)
Artin and Zhang: define projective scheme
P2q = Proj A := (tails(A),O, sh)
Artin, Tate and Van den Bergh:
A −→ (E, σ, L) −→ B = B(E, σ, L) and B = A/gA where g ∈ A3 is central
tails(A) −⊗AB −→ ←− (−)A tails(B) ˜ (−) −→ ←− Γ∗ coh(E) > i∗ < i∗
3. Moduli spaces for gk-2 modules
Describe A-modules M s.t. GKdim M = 2 Simplifications:
1. Assume that M is g-torsionfree
2. Assume that M is Cohen-Macaulay 3. Assume that M<0 = 0, M0 6= 0
4. The Hilbert series of M has the form hM(t) = e
(1 − t)2 − f
1 − t + g(t) where g(t) ∈ Z[t, t−1]. So fix e and f .
Note: e > 0, f ≥ 0.
A G(e, f ) cohP2q = Tails A G(e,f) Moduli spaces M ∈ G(e, f ) satisfies • H0(P2q,M(l)) = 0 for l < 0
• i∗M ∈ coh(E) finite dimensional length 3e Using Serre duality:
H2(P2q,M(l)) = Ext2(O, M(l))
∼
= Hom(M(l + 3), O)∗ = 0 for all l Using Euler form:
Via the derived equivalence of Bondal: M ∈ G(e, f ) ←→ M ∈ mod ∆ for which
• dimM = (2e + f, e + f, f ) • Hom∆(M, p) = 0 ∀p ∈ E
• Hom∆(p, M ) = 0 except finitely p ∈ E
In case of the Weyl algebra:
dimM = (2e + f, e + f, f ) and M (Z0), M (Z1) are surjective Using the relations of ∆:
Corresponds to pairs of matrices
{(X, Y ) ∈ M2e+f(C)2 | rk(Y X − XY − id) ≤ e}
for which, up to simulaneous conjugation in Gl2e+f(C), both X and Y are of the form
f e e e + f e ∗ ∗ 0 ∗ 0 0
Still have to sort out:
• For Sklyanin algebras
• properties of these spaces
• We expect for critical modules: smooth (affine??) varieties of dimension e2 + 1
4. Hilbert series of gk-2 modules
• Sufficient: determine Hilbert series of G(e, f )inv = critical objects in G(e, f ) • Necessary conditions (Ajitabh):
If M in G(e, f )inv then hM(t) = e (1 − t)2 − s(t) 1 − t where s(t) = P i siti ∈ Z[t] satisfies e > s0 > s1 > . . . ≥ 0 and P i si = f (1)
• In fact Ajitabh found necessary conditions for the appearing minimal resolutions
• s(t) satisfying (1) are represented by graphs in the form of a stair
Example:
e = 12 and s(t) = 9 + 6t + 5t2 + 4t3 + 1t6 The corresponding graph is
Counting parameters we showed that the converse is also true.
Theorem 1. There is a bijection
Hilbert functions of objects in G(e, f )inv l Polynomials s ∈ Z[t] s.t. e > s0 > s1 > . . . ≥ 0 and P i si = f given by h(t) = e (1 − t)2 − s(t) 1 − t
• Number of appearing Hilbert series is 2e−1 • We showed: Ajitabh’s necessary conditions
for minimal resulutions are also sufficient • Via Hilbert series: stratification of G(e, f )inv
there is a dimension formula for these strata unique stratum of G(e, f )inv with maximal
5. Presentation up to gk-1 modules grmod A = { f.g. right A-modules }
grmod A≤1 = {M ∈ grmod A, GKdim M ≤ 1}
Quotient map θ : grmod A → grmod A/ grmod A≤1 K, M ∈ grmod A are gk-1 equivalent if θ(K) ∼= θ(M )
Theorem 2 (Ajitabh and Van den Bergh). Every critical M ∈ G(e, f ) is gk-1 equivalent with a critical K ∈ G(e, 0) s.t.
0 → A(−1)e → Ae → K → 0
In the commutative case:
Every critical M ∈ G(e, f ) is gk-1 equivalent with a critical K ∈ G(e, e(e − 1)/2) s.t.
We proved that this is not the case for A. Example: e = 3 and
0 → A(−2)2 → A(−1) ⊕ A → M → 0 0 → A(−3) → A → K → 0
M has 10 parameters, K has 9.
Idea: only finitely many critical gk-1 modules to map M or K onto!
Corollary 1. Not every simple gk-1 over A1 is of the form A1/aA1, a ∈ A1.
Writing A1 = k[x, dxd ]:
Corollary 2. Not every system of differential equations in x can be reduced to a single equa-tion.